Presently, advances in network environments, such as the Internet, and in technology for compressing digital data has made it possible to obtain a large amount of content data such as music data and video data. Conventionally, using a search word indicating an attribute such as a characteristic and a property of the content has been used to retrieve content data desired by the user.
In the conventional system for retrieving content data using a search word, an attribute value indicating a content attribute is assigned to each content data. The search word is converted into information indicating the range of the attribute value associated with the search word. Then, pieces of content data specified by the attribute values included in the range of the attribute value are outputted as search results.
Such content data retrieving system as described above include a music retrieval system for retrieving music data using impression words (e.g. Japanese Laid-Open Patent Application No. 2004_-86189). An impression word is a qualitative word representing an impression on music content, e.g., “relaxing”.
According to the music retrieval system, an impression value is an index value representing an impression of the music set in advance for each music data. Moreover, an impression word inputted as a search word is converted to indicate the range of the corresponding impression value. The respective music data specified by the impression values included in the range of the impression value are retrieved as search results.
The music retrieval system is applied, for example, to car navigation apparatuses and PC music playback software.
With the conventional music retrieval system using impression words, however, the range of the impression value associated with an impression word is determined in a fixed manner. Therefore, in the case where the distribution of the impression values of respective pieces of music regarded as search targets is uneven, music is searched out only in the case of conducting a search using a specific impression word.
For example, in the case where a search target is a group of music obtained as a result of conducting a search using a specific artist's name, the impressions of many pieces of music are similar. That is to say that the absolute positions, within a space where impression values are presented, of the impression values of the respective pieces of music are close to each other. Therefore, in the case of conducting a search using an impression word, numerous pieces of music are selected for a specific impression word whereas no music is retrieved for an impression word other than that specific impression word.
Although a given impression word is the same, the range of the impression word intended by the user differs, e.g., the impression word “relaxing” used for the music performed by a rock artist who sings intensive music differs from the impression word “relaxing” used for the music performed by a female artist who sings calm music. This is because the user subconsciously knows that the possible range of impression values of music differs depending on an artist or a genre of the music, and thus represents, with an impression word, an impression relatively perceived in a group of music performed by the artist. However, the conventional music retrieval system using impression words does not take this point into consideration.
Thus, with the conventional music retrieval system using impression words, when the distribution of the impression values of the respective pieces of music regarded as search targets is uneven, the search results vary as well. As a result, the search results which reflect the user's intention cannot be outputted. In other words, it is not possible to narrow down the range of search targets, which is expected by the user.
The present invention is conceived in view of the above problem, and an object of the present invention is to provide a content data retrieval apparatus which retrieves, using a search word, the content data associated with an attribute value indicating a content attribute, and which appropriately narrows down the range of search targets even in the case where the distribution of attribute values for a group of contents regarded as search targets is uneven.